Revisit Agent Prompts

[neuromechanist]

Further testing of the agents show that the evaluation, faithfulness and conflict agents commonly lack catching semantic, meaning like grouping in parentheses. We should revisit the prompts to ensure that the additions in the annotation agent is properly tested and evaluated in other agents.

[neuromechanist]

Greatly improved in 52ddf46,

$ hedit annotate "reward_time: Reward delivery. The animal received a drop of juice or food for successful completion of the task (seconds)" --model anthropic/claude-haiku-4.5
Generating HED annotation (api mode)...
╭─────────────────────────────────────────────── HED Annotation - FAILED ────────────────────────────────────────────────╮
│ Annotation:                                                                                                            │
│   (Measurement-event, (Agent-action, ((Animal-agent, Animal/Primate), (Receive, Reward))), Duration/# s)               │
│                                                                                                                        │
│ Status: [red][ ] Valid[/] [yellow][ ] Faithful[/] [yellow][ ] Complete[/] [dim](5 validation attempts)[/]              │
│                                                                                                                        │
│ Errors:                                                                                                                │
│   - [PLACEHOLDER_INVALID] ERROR: [PLACEHOLDER_INVALID] "(Measurement-event, (Agent-action, ((Animal-agent,             │
│ Animal/Primate), (Receive, Reward))), Duration/# s)" has "#" placeholders, which are not allowed in this context.      │
│ (For more information on this HED error, see                                                                           │
│ https://hed-specification.readthedocs.io/en/latest/Appendix_B.html#placeholder-invalid.)                               │
│                                                                                                                        │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

Note: The reward_time: Reward delivery. at the beginning of the prompt still makes it difficult to handle, it might be removed for a better response.
Note2: The failure is likely due to HED validator not accepting standalone (#) in tags

The revised prompt in #62, result in:

$ hedit annotate "A monkey received a drop of juice or food for successful completion of the task" --model anthropic/claude-haiku-4.5
Generating HED annotation (api mode)...
╭─────────────────────────────────────────────── HED Annotation - SUCCESS ───────────────────────────────────────────────╮
│ Annotation:                                                                                                            │
│   Agent-action, ((Animal-agent, Animal/Monkey), (Perform, Task)), Feedback, (Reward, (Ingestible-object,               │
│ Gustatory-presentation))                                                                                               │
│                                                                                                                        │
│ Status: [green][x] Valid[/] [green][x] Faithful[/] [green][x] Complete[/] [dim](4 validation attempts)[/]              │
│                                                                                                                        │
│ Warnings:                                                                                                              │
│   - [TAG_EXTENDED] WARNING: [TAG_EXTENDED] Tag extensions found for [Animal/Monkey] in "Agent-action, ((Animal-agent,  │
│ Animal/Monkey), (Perform, Task)), Feedback, (Reward, (Ingestible-object, Gustatory-presentation))".  (For more         │
│ information on this HED warning, see https://hed-specification.readthedocs.io/en/latest/Appendix_B.html#tag-extended.) │
│                                                                                                                        │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

Note: The semantic grouping and capturing the sentence is significantly improved (see original instance in the first issue)